Integrated location fingerprinting and physical neighborhood for WLAN probabilistic localization

For the purpose of utilizing physical neighborhood relations of adjacent reference points (ARPs) in radio-map, a new approach by constructing both location fingerprinting database and physical neighborhood database in off-line phase is proposed to enhance the accuracy of wireless local area network (WLAN) probabilistic localization. In the on-line phase, we first rely on Bayesian inference to find the most adjacent points (MAPs) with respect to each testing point (TP). Then, based on the physical neighborhood database, we obtain the physical adjacent points (PAPs) corresponding to these MAPs. In the set of MAPs and PAPs, we choose the feature points (FPs) for the second Bayesian inference. Finally, we locate the TP at the geometric center of the chosen FPs which has the maximum posterior probabilities.